AI-Assisted Automatic Design Methods for Antennas and RF Circuits Based on Deep Learning

Radio-frequency (RF) antennas and circuits are essential components to applications such as wireless communication, satellite, radar, and the Internet of Things. Manual design process for antennas and circuits are dependent on expert knowledge that proves to be time-consuming and thus more efficient and autonomous design methods are preferred. This paper introduces two AI-assisted automatic design approaches employing deep learning (DL), specifically deep neural networks (DNNs) for antenna design and deep reinforcement learning (DRL) for circuit design. The DNNs predict antenna parameters and radiation gain, significantly reducing simulation time. RF coupler is adopted as the example for circuit design that utilizes DRL with a deep Q-network (DQN), dynamically selecting element types and parameter values to achieve desired power dividing ratios. The presented methods showcase outstanding efficiency and accuracy compared to traditional approaches, offering a significant advancement in automatic design processes for antennas and RF circuits.